Tuesday, February 05, 2008

pollsters: who's right?


The polls are all over the place in the various Super Tuesday states, perhaps most of all in the biggest prize, California (go figure), where Zogby has picked Obama by 12 and SUSA Hillary by 10. These, of course, run on the heels of January's primaries, 3 out of 4 having left the pollsters with egg on their face (in fact, the pollsters actually got worse as the season went on, getting NH more wrong than IA and SC more than all of them). Nevertheless, by comparing numbers we may at least get a sense of who's hitting closest to the mark so far.

Iowa: The first caucus finished Obama 38%, Edwards 30%, Clinton 29%. Here's how the big dogs pegged it:
ARG: Clinton 31, Obama 25, Edwards 21
Reuters/CSPAN/Zogby: Obama 31, Edwards 27, Clinton 24
CNN: Clinton 33, Obama 31, Edwards 22
McClatchy/MSNBC/Mason-Dixon: Edwards 24, Clinton 23, Obama 22

Oddly, I didn't find any polling from Rasmussen or SUSA near the primary day, so I'll assume they passed on it. Zogby is the only one who got the order right, but all the numbers are low, including his. Caucuses are notoriously difficult to poll, however, because people often end up supporting their 2nd or 3rd choices. And because caucuses are the stupidest, most irrational way to choose a candidate ever devised. Go democracy, and take individuality with you.

New Hampshire: As you remember, Clinton blew everyone's expectations here, including the pollsters, finishing 39 to Obama's 37 and Edwards' 17.
ARG: Obama 40, Clinton 31, Edwards 20
Zogby: Obama 42, Clinton 29, Edwards 17
Rasmussen: Obama 37, Clinton 30, Edwards 19
CNN: Obama 39, Clinton 30, Edwards 16
USA Today/Gallup: Obama 41, Clinton 28, Edwards 19

No one picked Clinton to win, but CNN, Rasmussen, and to a lesser extent ARG got Edwards' and Obama's numbers right. Zogby and Gallup overstated Obama's support. Looks like Rasmussen missed the least badly (but, ahem, still badly), only picking Obama to win by 7.

Nevada: another caucus, oy. After the voters were sufficiently separated by height and hair color, herded into their respective corners, shoed, tagged, asked to divide 783 by 4, given ceremonial wombat tooth necklaces, and made to burn incense for Gumle, the Great Woodpecker Spirit, they came up with Clinton 51%, Obama 45%, Edwards 4% (ouch!). The pollsters?
Zogby: Clinton 45, Obama 39, Edwards 6
Mason-Dixon: Clinton 41, Obama 32, Edwards 14
ARG: Clinton 45, Obama 32, Edwards 25
Research2000: Obama 32, Clinton 30, Edwards 27

Zogby was impressive in NV, nailing the order and the spread and even getting Edwards within the margin of error. ARG significantly overestimated Clinton's margin of victory (and Edwards' support), while Research2000 was the big loser of the night and I suspect may have accidentally polled Utah instead.

South Carolina: Obama apparently got tired of getting pushed around and cleaned Clinton's clock here, 55 to 27, with Edwards taking 18.
Zogby: Obama 41, Clinton 26, Edwards 19
ARG: Obama 39, Clinton 36, Edwards 22
SUSA: Obama 43, Clinton 30, Edwards 24
Rasmussen: Obama 43, Clinton 28, Edwards 17

By this time, someone at SUSA had finally turned on the news and found out there was a primary going on, but Mason-Dixon is still checked into the Reno Motel 6 with a bottle of Wild Turkey, traumatized by the horror of attempting to peg Obama support in a circus caucus, so they sat this one out. Everyone got the order, and everyone but ARG got Clinton's numbers right (they overestimated it again), and everyone but ARG and SUSA getting within 3 of Edwards' numbers. ARG really botched the spread, predicting a 3-point Obama win (it turned out to be 28) Zogby and Rasmussen got this one closest.

So what did we learn here, kids? Well, despite the great Zogby prediction snafu of 2004, Zogby and Rasmussen have done the best so far, with Rasmussen in fact taking top honors in both of the contests in which they polled. ARG, so far, has been probably the worst.

Zogby and Rasmussen both predict an Obama victory in California, despite other polls (most notably SUSA) who think Hillary will run away with it. It really only matters for media exposure and narratives since Dem primaries are mostly (if not all) proportional representation, but still, there's reason to be hopeful tonight.

No comments: